TRANSPORTATION RESEARCH RECORD llll 71

present standards of uniformity and level of illurninance and 3. M. G. Bassett, S. Dmitrevski, P. C. Kramer, and F. W. Jung. Mea­ luminance in fixed highway lighting cannot reveal a system's surements of Reflection Properties of Highway Pavement Samples. weakness. Journal of the Engineering Society, SepL 1981. 4. J.B. DeBoer and D. A. Schreuder. Theoretical Basis of Road Lighting Design. In Public Lighting, Chap. 3, Philips Technical Library, Eindhoven, Netherlands, 1967. REFERENCES 5. A. Erbay. Atlas of the Reflection Properties of Road Surfaces. Institut fuer Lichttechnik, Technische Universitaet Berlin, Berlin, 1. Determination of Minimum-Maximum Luminance Levels for Pro­ West Germany, 1974. grammable Lighting Operations. Internal Report Ministry of Transportation and Communications, Downsview, Ontario, Can­ ada, 1984. 2. CIB Committee TC-4.6. Calculation and Measurement of Lumi­ nance and Illuminance in Road Lighting. CIB Publication 30 (fC-4.6). Commission Internationale d'Eclairage, Paris, 1976. Publication of this paper sponsored by Committee on V1Sibility.

A Method of Calculating the Effective Intensity of Multiple-Flick Flashtube Signals

MARC B. MANDLER AND JOHN R. THACKER

A method of determining the effective intensity of light flashes A flashtube is a capacitive discharge device capable of emitting composed of multiple pulses (flicks) of light was devised. Detec­ brilliant flashes of light in extremely brief time periods (on the tion thresholds were measured for such flashes when the flick order of microseconds). The highly intense flashtube burst can frequency and flash duration were varied. Thresholds de­ be detected at great distances and has been noted as a conspic­ creased with increasing flick frequency and flash duration. At uous signal in a typical aid-to-navigation system (1). Moreover, each flick frequency the relationship between threshold and flash duration was well characterized by the Blondel-Rey rela­ the efficiency of converting input energy to visible output is tion (a = 0.2), provided a multiplicative frequency-dependent greater than that of an incandescent light (2). These factors fitting parameter was chosen. The fitting parameter, ~. in­ make the flashtube attractive as an aid to navigation. creased linearly with frequency between 5 and 20 Hz. A There are three major disadvantages associated with the use method of determining effective intensity was described that of flashtubes. The intense nature of the flick tends to momen­ uses the flick frequency, number of flicks, and the calculated tarily blind the close observer (3). Also, the duration of the effective Intensity of a single flick to arrive at the solution. It single flick is so brief that mariners have difficulty fixing the was concluded that this method should be used for all multiple­ flick signals, provided the single-flick duration is Jess than 0.01 exact location in the visual field (3 ). Finally, mariners report sec and the frequency is between 5 and 20 Hz. The method of difficulty judging the distance to the flashing source (3 ). The Allard should not be used, because it consistently overesti­ latter two difficulties can be ameliorated by presenting several mates effective intensity. flicks in rapid succession so that the appearance is not one of individual flicks, but of a longer-duration flash. Previous stud­ ies have shown that individuals can take line-of-sight bearings with greater speed and accuracy when the flash duration is increased in this way (4). U.S. Coast Guard Research and Development Center, Avery Point, The detection distance of a lighted aid to navigation is Groton, Conn. 06340-6096. valuable information because it not only allows one to calculate 72 TRANSPOKfATlON RF.sEARCH RECORD llll the range at which a light will become visible, but it is also one DEFINITIONS measure of the signal effectiveness of the aid. A typical aid-to­ navigation system is composed of both steady a.'l.d flashing A single light pulse from a flashtube is called a flick. When lights. The detection distance of a steady light, where intensity several flicks are presented in rapid succession so that dark does not vary with time, can be calculated from the familiar periods are not distinguished between the individual flicks, this Allard's law (5, p. 62): is referred to as a multiple-flick flash. The rate at which the flicks are delivered in a multiple-flick flash is termed the flick frequency. The flash duration or flash length is the time be­ tween onset of the first flick and the cessation of the last flick where and is a function of the flick frequency and the number of flicks. Figure 1 provides two examples of multiple-flick signals. E illuminance threshold of the eye [typically 0.67 sea- The bottom portion of Figure 1 shows relative intensity as a mile candle (6)], function of time for a 20-Hz, 13-flick signal. The time between I intensity of the light, each flick is 1 per flick frequency, which in this case is 0.05 sec. T transmissivity of the atmosphere, and The flash duration is 0.6 sec. ill the upper portion of the figure, D distance at which the light is visible. a 5-Hz, 4-flick signal is shown. Each flick is delivered every 0.2 sec. As with the 20-Ilz signal, the flash duration is 0.6 sec, As the intensity of a flashing light source is a function of time, though the number of flicks and total integrated light intensities the detection distance can be calculated by using Allard's law of the two signals are different. provided that a steady-light-equivalent intensity, termed the effective intensity of the light, can be determined. Effective intensity is defined as follows (7) :

5 Hz If a flash is found to be just seen in conditions in which a steady light of intensity le is also just seen at the same distance and in the same atmospheric conditions, the flash is said to have an effective intensity le. 20 Hz The three generally accepted methods of calculating the effec­ .?• i tive intensity of a single flash are the methods of Allard (5) (not 'ii a: to be confused with Allard's law discussed previously), 0 Schmidt-Clausen (8, 9), and Blondel-Rey-Douglas (10, 11). All 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 these methods are condensed and described by the futemational Time (seconds) Association of Lighthouse Authorities (IALA) (7). FIGURE 1 Plot of intensity versus time of As noted earlier, the increased duration of the multiple-flick multiple-flick flashes. flash is desirable, but to incorporate these multiple-flick flash­ tube devices in an aid-to-navigation system requires a method of specifying its detection range. The IALA formally recog­ nized that of the three single-flash methods of calculating APPARATUS effective intensity, only the Allard method is appropriate for calculating the effective intensity of multiple-flick flashes (12): Figure 2 shows a schematic of the apparatus used for data collection. A flashtube (Automatic Power, Inc., Part The reason for recommending ... the method of Allard for 9001-0295) was installed inside an integrating sphere. A 0.025- trains of rapidly repeated fl ashes was that this method would in. aperture was attached to the output of the integrating sphere yield an effective intensity that increased with increasing num­ and defined the source size. A variable neutral-density wedge ber of flashes in the train and approached asymptotically a steady-state response that for very rapid rates was identical with provided computer control of illuminance of the signal. The Talbot's Law. The other two methods could not yield any signal was reflected off a mirror and superimposed on a low­ satisfactory effective intensity for trains of flashes. It cannot, luminance (0.0045-ft-lambert) background provided by a however, be said that there is any direct experimental confirma­ Kodak slide projector. tion of the effective intensity obtained by the Allard method for The signal appeared as a point source (actual angular diame­ repeated flashes. ter of 0.0003 mrad) when viewed from 83.25 in. The back­ ground subtended approximately 1.0 rad by 1.0 rad. To mini­ The Allard method involves lengthy computer calculations of mize the observer's uncertainty of the position of the signal, the explicit solution of a differential equation (7). The primary four small low-intensity fixation points were provided, each purpose of this work was to find a simple, accurate method of 17.5 mrad from the signal location. determining the effective intensity of multiple-flick signals The fiashtube circuitry was modified so that it could be based on the characteristics of the signal. A secondary goal of triggered by a computer pulse. A single output flick was pre­ this work was to provide the missing experimental confirma­ sented for each input trigger with the maximum rate being 50 tion of the Allard method. flicks per second Mandler and Thack£r 73

Slide projector Variable density! wedge.

Integrating sphere ~ "-. Flxation plate /

Viewing Aperture with lour surrounding Observer fixation lights

FIGURE 2 Schematic of laboratory apparatus.

CALIIlRATION Figure 3 shows that the peak output occurs about 20 µsec after flick onset and the intensity decays to 10 percent of peak Figure 3 is a plot of the intensity as a function of time for a after 55 µsec, with negligible low-level output for as long as single near-threshold flick. To obtain this curve, an EG&G about 85 µsec after onset. photomultiplier tube (PMT) with photometric filter was posi­ tioned in the apparatus where the observer's eye was typically positioned. The PMT, which had a rise time of less than IO PROCEDURE nsec, was calibrated against an EG&G Model 555 photometer system using a steady light, so that the relationship between Detection thresholds were obtained for a total of 34 separate illuminance and output voltage from the PMT was known. signals. Table 1 shows the six flick frequencies that were used,

02

IJ ' I ~ .015 I I '\ ro I Qi TI c ro J " ~ I "' , 01 ·u;~ ' c I cQ) I ' I , I ' \ .005 '- J ' I'... I \. I '\. , ...... 0 / - 0 10 20 30 40 50 60 70 80- 90 100 Time (Microsecond) FIGURE 3 Intensity profile of a single flick. 74 TRANSPORTATION RESEARCH RECORD llll

TABLE 1 TEST SIGNALS Flick Frequency (Hz) No. of Flicks Flash Duration (sec) 5 1,2,3,4 0.000,0.200,0.400,0.600 8 1,2,4,6 0.000,0.125,0.375,0.625 11 1,3,5,7,9 0.000,0.182,0.363,0.546,0. 727 14 1,3,5,7,9,11 0.000,0.143,0.286,0.429,0.572,0. 714 17 1,3,5,7,9,11,13 0.000,0.118,0.235,0.353,0.471,0.588,0.706 20 1,3,5,7,9,ll,13,15 o.ooo,o.100,o.200,o.300,o.400,0.500,0.600,0. 100

the number of flicks provided at each frequency, and the corre­ RESULTS sponding flash duration (see Equation 4). A "staircase" procedure was used to measure thresholds Figure 4 shows one observer's thresholds as a function of flash simultaneously for all the signals of a parlicular frequency. On duration for six flick frequencies. Flash duration is defined to a given trial, one of the 34 signals was presented and the be the time from the start of the first flick to the end of the last observers responded as to whether or not the signal was de­ flick. Threshold is defined as the illuminance of the signal that tected by pressing one of two computer-readable switches. If could be detected 79 percent of the time. Because the illumi­ the signal was not detected, the illuminance was raised by 0.1 nance of each flick varies with time, the peak illuminance is log unit (25.9 percent). When the signal was detected three used as the measure of threshold. consecutive times, the illuminance was decreased by 0.1 log For all frequencies, as flash duration increased, lower peak unit. This illuminance staircase continued until the illuminance illuminance was required to detect the light, and thus the had reversed direction eight times. Threshold was taken to be threshold is considered to have decreased. Moreover, as fre­ the mean of the peak illuminances where the staircase reversed quency increased, and thus the number of flicks per flash direction. This procedure yields a threshold that corresponds to increased, thresholds also decreased. approximately a 79 percent probability of detection (13 ). As in any study of visual sensitivity, observers have different Four observers with 20/20 vision or better wearing spectacles thresholds (14). In this experiment, the thresholds of the most participated in this experiment. Observers dark adapted for at sensitive and least sensitive observers differed by more than a least 20 min before data collection began. All viewing was factor of 2. Because concern here is not with an absolute done monocularly (with one eye), primarily because of appa­ measure of threshold, but rather a relative measure of how ratus limitations. threshold changed with flash duration and frequency, all ob­ The test signals were always provided in the center of the server one-flick thresholds (flash duration = 80 µsec) were four-light fixation target (see Figure 2) and observers were normalized to 1.0. This means that thresholds at all other flash required to fixate at this point throughout the experiment. durations were proportionately less than 1.0, revealing the ex-

2500

2300

2 100

1900 0 E ~ 1700 "O 0 .c V) 1500 5 Hz ~ .c I- 1300 8 Hz

1100

11 Hz

14 Hz 700 17 Hi 20 Hz

500 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ .1 .4 .5 .6 .7 .B 9 Flash Leng1h (Seconds) FIGURE 4 Observer thresholds as a function of flash duration. Mandler and Thacker 75 tent to which the illuminance of these other signals could be circles show the predictions of the Allard integral, the preferred reduced, relative to ·the one-flick threshold, to bring it to the method of calculating effective intensity of multiple-flick sig­ threshold criterion. Moreover, because the authors' interest was nals (7). with effective intensity and not threshold, the data were con­ It is clear that the Allard method overestimates the effective verted from a threshold ordinate to an effective intensity ordi­ intensity of the multiple-flick signals. The amount by which nate. Because two signals at threshold have identical effective effective intensity is overestimated increases with frequency intensities, the data of Figure 4 represent the peak illuminances and flash duration. For example, at 20 Hz the Allard method of various test signals of equal effective intensities. The thresh­ overestimates effective intensity by as much as 22 percent. old ordinate of Figure 4 can be converted to an effective The solid curves through the data are best fits of the function intensity ordinate by taking the reciprocal of threshold. That is, if one signal has a threshold that is 0.5 times that of another le = 1.0 + p * [t/(a + t)] (1) signal, then it has twice the effective intensity. Figure 5 shows the relative effective intensity functions of where the six test frequencies. The squares represent the mean of the four observers, and the vertical bars are ±1.0 standard error of P = fitting parameter, the mean. The solid curve through the data is a theoretical flash duration, and function fit to the data and will be discussed in detail later. The a constant of 0.2.

5 HI SHI 5 s

>- ·~ 4 4 ~ ~ .5 3 'i~ 3 = w 2 2 ~ 0 Ia: 1~ , ..-----r--

Oo .1 2 .3 .4 .s .6 .7 8 .9 Oo .2 3 ,4 s .6 7 .8 9 Flash Length (Seconds) Flash Length (Seconds)

11 Hz 14 Hz 5 5

f 4 4 .s~ 3 • • • . • • • 'i~ 3 = w 2 2 ~ ""' a:11 1

00 0.0 .1 .2 3 4 .5 .6 7 .8 .9 . I .2 .3 .4 .5 .6 .7 .8 9 Flash Length (Seconds) Flash Length (Seconds)

• 0 • 4 • • 0 0 000000 3

0 o .1 .2 .3 .4 .5 .6 .1 a .9 ° ~0-~-.,,2--:_ 3=---_..,.4--=_ 5=--""". s=---.-=-1-""'. s-""". 9=--__. Flash Length (Seconds) Flasn Length (Seconds) FIGURE 5 Relative effective intensity functions. 76 TRANSPORTATION RESEARCH RECORD 1111

o Observer data o Allard predictions 4

3

2

0 5 10 15 20 Flick Frequency (Hz) FIGURE 6 Best-fitting ~.a = 0.20.

This is essentially the Blondcl-Rey relation (JO). The constant Proposed Method of Calculating Effective Intensity 1.0 was added in keeping with the normalization performed on the data. By optimizing ~. it was possible to fit the group data at For any multiple-flick signal the effective intensity can be each frequency with this function, as shown by the solid curves determined by using Figure 5 as a nomograrn and finding, of Figure 5. along the ordinate, the value of the relative effective intensity The 13 providing the best fit at each frequency is shown in for the appropriate frequency and flash duration. This value, Figure 6. The observer data are shown as squares and the when multiplied by the calculated effective intensity of a single Allard calculation as circles. The lines are least-squares fits to flick, yields the effective intensity of the signal. the data. An alternative, but equivalent, approach is to derive the As frequency increases, the fitted 13 increases. Again, as equations that can be used to calculate the effective intensity. noted earlier, the fits for the Allard predictions are greater than Equation 1, which was used to fit the observer data of Figure 5, those for the observer data. The 13 fits for the Allard predictions provides the relative effective intensity of any signal provided fall along a straight line with slope of 0.243 and intercept of the fitting parameter, 13, is known. The effective intensity (/e) of -0.488. The observer data can be reasonably approximated by a multiple-flick flash can be determined from a straight line with slope of 0.203 and intercept of -0.577. le = Ie1 * 1.0 + f3 [t/(a + t)] (2)

DISCUSSION OF RESULTS /el is the effective intensity of a single flick. To calculate this single-flick effective intensity, any of the three methods pro­ The purpose of this experiment was to establish a method of posed by IALA is sufficient (7). In these calculations the three specifying the effective intensity of multiple-flick flashes. It methods yielded nearly identical results for the flick shown in has been shown that for each flick frequency, observer data can Figure 3. The calculated effective intensities from these three be fit with a Blondel-Rey equivalent function (Figure 5) having methods are as follows: a frequency-dependent fitting parameter, 13 (Figure 6). Had the thresholds been independent of frequency, the ratio of the Method Effective lnlem·ity effective intensities at a particular flash duration would have 1 Schmidt-Clausen 4.6189 x 10· been the same as the ratio of the total energy in the flashes Blondel-Rey-Douglas 4.6170 x 10-1 7 regardless of frequency. This is not the case, as can be seen in Allard 4.6800 x 10" Figure 5. A 13-flick 20-Hz signal and a 4-flick 5-Hz signal both have flash durations of 0.600 sec. The ratio of the total energies The value 13 can be found from the least-squares fit in Figure 6, of these two signals is 3.25, yet the ratio of the effective given the flick frequency. It can be calculated by intensities is 2.30. Consequently, a frequency-dependent rela­ tionship is required, as shown in Figure 6. 13 = (0.203 * f) - 0.577 (3) Mandler and Thacker 77

TABLE 2 RELATIVE EFFECTIVE INTENSITIES OF COMMON FLASHTUBE SIGNALS Flick Frequency No. of Flicks per Flash (Hz) 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 5 1.219 1.292 1.329 1.350 6 1.291 1.401 1.458 1.493 1.517 7 1.352 1.497 1.575 1.625 1.659 1.684 8 1.403 1.582 1.683 1.748 1.793 1.827 1.852 9 1.447 1.658 1.781 1.862 1.919 1.961 1.994 2.020 10 1.485 1.727 1.872 1.969 2.038 2.090 2.130 2.162 2.189 11 1.518 1.789 1.956 2.068 2.150 2.212 2.260 2.299 2.331 2.357 12 1.547 1.845 2.033 2.162 2.256 2.327 2.384 2.430 2.468 2.499 2.526 13 1.573 1.897 2.105 2.250 2.357 2.437 2.504 2556 2.600 2.637 2.668 2.695 14 l.597 1.944 2.172 2.333 2.452 2544 2.618 2.679 2.728 2.770 2.805 2.837 2.864 15 1.612 1.989 2.234 2.411 2.543 2.645 2.728 2.795 2.851 2.899 2.939 2.974 3.005 3.033 16 1.637 2.028 2.293 2.484 2.629 2.742 2.833 2.908 2.971 3.024 3.069 3.109 3.143 3.174 3.201 17 1.654 2.065 2.348 2.554 2.711 2.835 2.935 3.017 3.086 3.145 3.195 3.240 3.278 3.312 3.342 3.370 18 l .670 2.100 2.399 2.620 2.789 2.923 3.032 3.122 3.198 3.262 3.318 3.367 3.410 3.448 3.482 3.512 3.539 19 1.684 2.132 2.448 2.682 2.864 3.008 3.126 3.224 3.306 3.377 3.438 3.491 3.538 3.580 3.617 3.651 3.681 3.708

where f is frequency in hertz. Flash duration (t) is calculated by intensity can be used for any multiple-pulse light flash as long as each pulse is less than 0.01 sec. t = (l/J) * (n - 1) + d (4) Visual Time Constant (a) where

It has been internationally agreed (7) that for nighttime obser­ f = frequency (Hz), vation the visual time constant, a, used for calculations of n = number of flicks in the flash, and effective intensity, should be equal to 0.2. Nighttime observa~ d = duration of a single flick. tion is assumed to be at a background luminance less than 0.1 cd/m2 (7). The background used in this experiment was a.ors This equation simply calculates the time between the start of 2 the first flick and the end of the last flick. cd/m , considerably Jess than the maximum penniucd. For the Table 2 provides the solution of the foregoing equations for curve fitting performed on the data of Figure 5, a was assumed flick frequencies between 5 and 20 Hz and various numbers of to be 0.2. It can be seen that the theoretical curve does an flicks. The effective intensity can be determined by multiplying adequate job of fitting the observer data. It was disturbing, the appropriate tabulated value in Table 2 by the effective though, that the Allard method overpredicted le. As an ex­ ercise, the analysis was repeated using different values of a in intensity of a single flick. search of an a that would bring the observer data and the Allard calculation into agreement. The observer data were refitted Flick Duration with different values of a and the Allard calculation was per­ formed with these same values and the results were compared. The method proposed here is based on threshold measurements Figure 7 shows the best-fitting ~ for the observer data and the for signals with the 80-µsec flick profile of Figure 3. Not all Allard calculation assuming an a of 0.155. The least-squares fit flicks or light pulses have the same temporal profile, and thus to the observer data yields a slope of 0.189 and an intercept of the generality of this method for other flick profiles must be -0.540, whereas the slope and intercept for the Allard calcula­ addressed. It is well established that the shape of a light pulse tions are 0.183 and -0.445, respectively. does not affect threshold or effective intensity if its duration is less than the critical duration (i.e., Bloch's law is obeyed) (10, 15). Within certain time intervals, the eye acts as a perfect Areas of Further Investigation integrator, and thus pulse shape is irrelevant. Although the time period over which Bloch's law is obeyed varies with many In this experiment only a single background luminance was stimulus conditions (16, p. 154), it can conservatively be esti­ used. It is of interest to determine whether the slope of the mated at 0.01 sec. It appears that as long as the flick duration is curve in Figure 6 varies with background luminance so that the less than 0.01 sec, this method can be used. Although this has generality of this approach can be assessed. Further, it was not been verified empirically in this work, two considerations argued that the proposed method can be used with many dif­ support this conclusion. First, the effective intensity, as calcu­ ferent flick profiles and durations. This argument should be lated by any of the three methods of IALA (7), is independent empirically verified. of pulse shape for pulses between 0.0 and 0.01 sec. Second, the Another area-in which much work is needed is in determin­ amount by which Allard's method overestimates effective in­ ing the optimum flick frequency and flash duration. Such an tensity is constant provided the flick duration is less than 0.01 approach not only must take into account the effective intensity sec. It is concluded that this method of calculating effective of the signal, but also must be concerned with the speed and 78 TRANSPORTATION RESEARCH RECORD 1111

o Observer data o Allard predictions 4

cc.

0 5 10 15 20 Flick Frequency (Hz) FIGURE 7 Best-fitting ~. a = 0.155. accuracy with which a bearing can be taken and must perform a reported here do not fully address the issue of the optimum comparative analysis of battery power requirements for such flashtube signal to incorporate into field-deployable hardware. signals. Thacker (4) showed that as flash duration increases, There remain several issues that must be resolved before the speed and accuracy in taking a bearing improve. Montonye and optimum signal is chosen. Accordingly, the following recom­ Clark (2) performed a partial analysis of flashtube battery mendations are made: power requirements. Edgerton (17) has measured the relation­ ships between the size of the flashtube storage capacitor, initial 1. Research should be conducted to specify the flashtube capacitor charging voltage, the peak output intensity, the elec­ energy consumption as a function of number of flicks em­ trical input to visual output efficiency, and the flick duration. ployed in a flash and the circuitry of the system. This analysis Before flashtubes are widely deployed in the field, all these would compare and contrast the results obtained from standard approaches to flashtube optimization must be fully analyzed in incandescent sources. This work will reveal how signals are conjunction with one another. limited by energy consumption. Work should also be done to optimize the flashtube circuitry for Coast Guard applications. 2. Research should be sponsored that addresses issues of CONCLUSIONS human performance with respect to flashtube signals. Specifi­ cally, problems of depth perception and difficulty in obtaining a TI1e following conclusions are based on the resuits reported fix on the iight should be studied more carefuliy. here. 3. Once the foregoing data have been collected, an optimum range of flashtube signals should be selected based on perfor­ 1. The proposed method of calculating effective intensity mance measures, calculated effective intensity, and energy c6n­ (le) should be used for any multiple-flick signal regardless of sumption considerations. the single-flick time-intensity profile, provided the single-flick duration is less than 0.01 sec and the flick frequency is between 5 and 20 Hz. The effecti;ve intensity can be calculated from REFERENCES Equations 2, 3, and 4 found in the discussion section. 2. The Allard method should not be used for calculating the 1. C. B. Devoe and C. N. Abernethy. Field Evaluation ofLocomotive effective intensity of multiple-flick flashes because it consis­ Conspicuity Lights. Report FRA-OR-D-75-84. Federal Railroad Administration, U.S. Department of Transportation, 1975. tently overestimates the effective intensity. The greater the flick 2. J. T. Monton ye and G. P. Clark. Evaluation ofLS-59 Xenon Flash­ frequency, the greater the error. tube System. U.S. Coast Guard, 1970. 3. D. F. Murphy. A Laboratory Evaluation of the Suitability of a Xenon Flashtube Signal as an Aid-to-Navigation. Master's thesis. RECOMMENDATIONS Naval Postgraduate School, Monterey, Calif., 1981. 4. J. Thacker. An Evaluation of Flashtube Signal Characteristics. Report CG-D-26-84. U.S. Coast Guard, 1984. The purpose of this effort was to determine a means of calculat­ 5. E. Allard. Memoire sur l'lntensite et la Portee des Phares. Im­ ing the effective intensity of a multiple-flick signal. The results primerie Nationale, Paris, France, 1876. TRANSPORTATION RESEARCH RECORD llll 79

6. Visual Signalling, Theory and Application lo Aids-to-Navigation. 12. fo.ternational Association of Lighthouse Authorities. Recommen­ Ocean Engineering Report 37, CG-250 Series. U.S. Coast Guard, dations for the Calculation of Effective Intensity of a Rhythmic 1970. Light. IALA Bulletin, Vol. 2, 1981, p. 27. 7. International Association of Lighthouse Authorities. Recommen­ 13. G. B. Wetherill and H. Levitt. Sequential Estimation of Points on a dations on the Determination of the Luminous Intensity of a Psychometric Function. British Journal of Mathematical and Sta­ Marine Aid-to-Navigation Light. JALA Bulletin, Vol. 3, 1978, p. 2. tistical Psychology, Vol. 18, 1965, p. 1. 8. H. J. Schmidt-Clausen. Experimental Investigations of the Valid­ 14. S. Hecht, S. Shlaer, and M. H. Pirenne. Energy, Quanta, and ity of the Blondel-Rey Equation. Report 5-1-1. Presented at Eighth Vision. Journal of General Physiology, Vol. 25, 1942, p. 819. International Conference on Lighthouses and Other Aids to Navi­ 15. A. M. Bloch. Experiences sur la Vision. Paris: Soc. Biol. Mem., gation, Stockholm, Sweden, 1970. Vol. 37, 1885, p. 493. 9. H. J. Schmidt-Clausen. The Influence of the Angular Size, Adap­ 16. N. R. Bartlett. Thresholds as Dependent on Some Energy Rela­ tation Luminance, Pulse Shape and Light Colour on the Blondel­ tions and Characteristics of the Subject. In Vision and Visual Rey Constant "a." In The Perception and Application of Flashing Perception (C. H. Graham, ed.), John Wiley and Sons, New York, Lights, London, Adam Hilger, 1971. 1965. 10. A. Blonde! and J. Rey. Application aux Signaux de la Loi de 17. H. E. Edgerton. Electronic Flash, Strobe, 2nd ed. MIT Press, Perception des Lumieres Breves a la Limite de Leur Portee. Cambridge, Mass., 1979. Journal de Physique, I (5th Series), 1911, p. 643. 11. C. A. Douglas. Computation of the Effective Intensity of Flashing Lights. Illuminating Engineering, Vol. 52, 1957, p. 641. Publication of this paper sponsored by Committee on Visibility.

Evaluation of Alternative Sign-Lighting Systems To Reduce Operating and Maintenance Costs

JONATHAN E. UPCHURCH AND JEFFREY T. BORDIN

The study objective was to identify a sign-lighting system that monly used fluorescent system, it uses one-third as much elec­ has a lower electric power cost and reduced maintenance tric power and has about one-third of the annual owning and requirements and that provides adequately for the motorists' operating costs. The recommended system has a satisfactory needs in terms of legibility and illumination level. Twenty-five illumination level and provides the best legibility distance of candidate lighting systems were identified through a review of the 10 systems tested. technical data and specifications for lamps and fixtures by an independent lighting expert. Photometric tests and computer analyses of sign illumination levels reduced the number of During the past 5 years interest has been increasing nationwide candidates to 10 alternative systems, which were then field tested. Each alternative lighting system was field tested for 10 in overhead guide-sign lighting because of the increasing cost to 14 months. Sign luminance was measured with a tele­ of the energy to provide illumination. In California, for exam­ photometer. Power consumption was monitored. Maintenance ple, the armual cost of electric power to illuminate overhead requirements and lamp life were noted. A human factors study signs on the freeway system increased from $993,000 in FY determined legibility distance and rated viewing comfort, 1977-1978 to $2,200,000 in FY 1982-1983 (W.A.J. Hoverstern, lighting uniformity, and color rendition. An economic analysis California Department of Transportation, unpublished data, was performed in which the initial cost of acquiring and in­ June 1985). The nationwide cost of power for overhead sign stalling the lighting systems and annual costs for electric power, washing, relamping, and ballast replacement were con­ lighting (for all overhead signs on all roadway systetns) was sidered. A lighting system using the high-pressure sodium light estimated to be about $20 million armually in 1986. source was recommended. Compared with the existing com- In addition to the cost of electric power, highway agencies are also concerned about the maintenance costs and labor Department of Civil Engineering, Arizona State University, Tempe, requirements for sign-lighting systems. California's annual Ariz. 85287. maintenance cost for its overhead signs is $800,000 per year